QuantumGenome EDS is a real-data genomics discovery project that converts variant data into an Ising-style interaction model and uses QAOA-based search to identify low-energy, disease-loaded multi-variant patterns.
The project began as a quantum-assisted Ehlers-Danlos / collagen discovery pipeline built from ClinVar and later expanded into a Jetson-first local discovery framework with same-gene, same-disease, position-window, mixed-biologic, novel-candidate, and epistasis-hypothesis phases.
This project has documented end-to-end validation on real data.
- Parsed 4,129,628 ClinVar variants and built disease-specific EDS and collagen panels.
- Validated h[i] scoring from ClinVar significance and J[i,j] scoring from shared disease annotations.
- In a PLOD1 baseline run, the lowest-energy bitstring was
0001at-5.0, correctly isolating the disease-driving variant. - In a B3GALT6 EDS cluster run, four pathogenic / likely pathogenic variants sharing spondylodysplastic EDS annotation produced pairwise
J[i,j] = +1.2and a lowest-energy bitstring of1101at-4.0. - In a 1000 Genomes chr22 pilot, the full real-data loop was validated: genotype matrix -> h/J -> classical energy spectrum -> QAOA simulator -> measured bitstrings.
This public repo includes runnable Python scripts with placeholder IBM settings so another user can:
- copy
config_template.pytoconfig.py - enter their own IBM credentials and backend
- run the ETL / panel / QAOA workflow themselves
The file quantum_model_builder.py includes:
- local Aer simulator mode
- IBM Quantum submission mode through
QiskitRuntimeService - placeholder settings in
config_template.py USE_IBM = Falseby default for safe first runs
To use IBM Quantum:
- copy
config_template.pytoconfig.py - fill in your own
IBM_TOKEN,IBM_INSTANCE, andIBM_BACKEND - set
USE_IBM = Trueinquantum_model_builder.py - run
python3 quantum_model_builder.py
This repository is publicly visible for portfolio, review, and discussion purposes only. All rights are reserved by the author unless explicit written permission is granted. No open-source license is provided for this repository.